Abstract
IMPORTANCE: In an aging and increasingly complex patient population, risk estimation tools are critical for optimizing clinical decision-making. The Clinical Frailty Scale (CFS) evaluates frailty and mortality risk, whereas the Good Outcome Following Attempted Resuscitation (GO-FAR) score aims to estimate survival after potential in-hospital cardiac arrest. However, their ability to estimate all-cause mortality in medical inpatients across all ages remains unclear. OBJECTIVE: To assess the accuracy of the CFS and GO-FAR score in estimating long-term all-cause mortality among medical inpatients. DESIGN, SETTING, AND PARTICIPANTS: This ancillary analysis of the CLEAR trial, a cluster randomized trial, included 2840 medical inpatients from 6 Swiss teaching hospitals. INTERVENTION: At admission from June 1, 2019, to April 30, 2023, the CFS, GO-FAR, and Charlson Comorbidity Index (CCI) scores were recorded. Regardless of randomization group, long-term follow-up of vital status occurred between June and October 2024. MAIN OUTCOMES AND MEASURES: The outcome was long-term all-cause mortality. RESULTS: Among 2840 patients (mean [SD] age, 68.5 [15.8] years; 1552 [54.6%] male), 969 (34.1%) died during a mean (SD) follow-up of 3.3 (0.91) years. The GO-FAR and CFS showed good discriminatory performance for all-cause mortality with time-dependent areas under the receiver operating characteristic curve (AUROC) of 0.78 and 0.74, respectively. Combining both scores with the CCI in a combined regression model improved the overall AUROC to 0.87. Patients in the highest-risk categories had significantly increased all-cause mortality risk (GO-FAR: adjusted hazard ratio, 20.31; 95% CI, 14.71-28.06; P < .001; and CFS: adjusted hazard ratio, 8.69; 95% CI, 6.83-11.07; P < .001). Both scores demonstrated high specificity in their highest-risk categories (GO-FAR: 98.2%; CFS: 99.1%) but low sensitivity (GO-FAR: 7.8%; CFS: 9.4%). Results indicated that of 100 hospitalized patients in the highest GO-FAR category (≥14 points), approximately 69 are expected to die, whereas among those in the lower categories, approximately 67 are expected to survive (positive and negative predictive value of 69.1% and 67.3%, respectively). CONCLUSIONS AND RELEVANCE: In this ancillary study of the CLEAR trial, the GO-FAR score and CFS effectively estimated long-term all-cause mortality, particularly when combined with the CCI. These tools may help clinical decision-making, resource allocation, and advanced care planning in aging, multimorbid patient populations.